In a clinical study of $n=50$ patients, the Wald statistic for testing $H_0: \theta=0$ yields a value of $W=3.84$. If the likelihood surface is highly skewed such that the quadratic approximation of the Wald method is invalid, which of the following is the most robust alternative to assess significance?